Virtified Tech Digest (29 October 2026)
In this edition of Virtified's Tech Digest, we see multiple security announcements—for both traditional hypervisor environments and for agentic infrastructure. Virtified expects microVMs to become a hot topic in AI infrastructure and recent announcements from Google and Microsoft showcase this.
From Virtified
15 September 2026 | Michael Warrilow | Virtified
Michael Warrilow critiques the 2026 Gartner Magic Quadrant for Server Virtualisation Platforms, assessing each vendor's position as overrated, about right or underrated.
Virtified in the news
26 October 2026 | Simon Sharwood | The Register
A Tesco job advertisement suggests the retailer plans to migrate its in-store edge virtualisation from VMware to Proxmox, while its contract lawsuit against VMware continues.
Virtual security
19 October 2026 | Simon Sharwood | The Register
Researcher Paulos Yibelo's PwnEscape, a critical Linux KVM flaw, lets a guest VM reach the host using malformed IPv4 packets; an upstream kernel fix is now flowing into major distributions.
13 October 2026 | Prashanth Shenoy | VMware Cloud Foundation (VCF) Blog
VMware announces HyperArmor, a protective layer to ESXi hosts. It combines real-time virtual patching, allow-only-trusted-process guardrails and MFA for SSH logins to counter AI-driven exploits.
Expected general availability is early 2027.
MicroVMs
22 October 2026 | Nirav Mehta, Michael Bachman and Kevin Hutchins | Google Cloud
Google Cloud announced Agent Compute Engine, a hardware-isolated microVM service for AI agents coming to preview, and extended its Gemini for business agent governance to multicloud, on-premises and endpoint environments.
22 October 2026 | Ramanathan Venkatapathy and Rohit Jnagal | Google Cloud
Google engineers explain how Agent Compute Engine runs AI agents in hardware-isolated microVMs that hold no credentials, idle at near-zero cost, share deduplicated memory and keep persistent workspaces between sessions.
7 October 2026 | Logan Iyer | Windows Developer Blog
Microsoft Execution Containers, now generally available, confine AI agents to developer- or IT-approved files and network access, using sandboxes on Windows, macOS and Linux; Intune and Entra integration is coming.